[llring problem]: Unable to add packets dequeue from llring to packetbatch

[DSPhnb]

We want to implement multi-queue scheduling based on the Queue() module(core/modules/queue.cc), so we created multiple llring(core/kmod/llring.h) type queues. We want a specific number of packets from the queue to be dequeued and added to processbatch. Doing the following in runtask() will throw an error:

The places I modified are marked

struct task_result Queue::RunTask(Context *ctx, bess::PacketBatch *batch,
                                  void *) {
  if (children_overload_ > 0) {
    return {
        .block = true,
        .packets = 0,
        .bits = 0,
    };
  }

  // const int burst = ACCESS_ONCE(burst_);
  const int pkt_overhead = 24;

  uint64_t total_bytes = 0;
  int ret;
  uint32_t cnt=0;

  // My Modification
  // uint32_t cnt = llring_sc_dequeue_burst(queue_, (void **)batch->pkts(), burst);

  bess::Packet *pkt;
  ret = llring_sc_dequeue(queue_, (void **)&pkt);       //queue type: struct llring
  if(!ret){
    batch->add(pkt);
    cnt++;
  }

  if (cnt == 0) {
    return {.block = true, .packets = 0, .bits = 0};
  }

  stats_.dequeued += cnt;
  batch->set_cnt(cnt);

  if (prefetch_) {
    for (uint32_t i = 0; i < cnt; i++) {
      total_bytes += batch->pkts()[i]->total_len();
      rte_prefetch0(batch->pkts()[i]->head_data());
    }
  } else {
    for (uint32_t i = 0; i < cnt; i++) {
      total_bytes += batch->pkts()[i]->total_len();
    }
  }

  RunNextModule(ctx, batch);

  if (backpressure_ && llring_count(queue_) < low_water_) {
    SignalUnderload();
  }

  return {.block = false,
          .packets = cnt,
          .bits = (total_bytes + cnt * pkt_overhead) * 8};
}

CommandResponse Queue::CommandSetBurst(
    const bess::pb::QueueCommandSetBurstArg &arg) {
  uint64_t burst = arg.burst();

  if (burst > bess::PacketBatch::kMaxBurst) {
    return CommandFailure(EINVAL, "burst size must be [0,%zu]",
                          bess::PacketBatch::kMaxBurst);
  }

  burst_ = burst;
  return CommandSuccess();
}

Modified as follows

// my modification
  // uint32_t cnt = llring_sc_dequeue_burst(queue_, (void **)batch->pkts(), burst);

  bess::Packet *pkt;
  ret = llring_sc_dequeue(queue_, (void **)&pkt);       //queue type: struct llring
  if(!ret){
    batch->add(pkt);
    cnt++;
  }

The error displayed by bessed.ERROR is as follows:

Log file created at: 2022/09/20 01:09:45
Running on machine: ubuntu
Log line format: [IWEF]mmdd hh:mm:ss.uuuuuu threadid file:line] msg
F0920 01:09:45.085120  9286 debug.cc:405] A critical error has occured. Aborting...
Signal: 11 (Segmentation fault), si_code: 1 (SEGV_MAPERR: address not mapped to object)
pid: 9270, tid: 9286, address: 0x7f16c9225838, IP: 0x557a0a5494a7
Backtrace (recent calls first) ---
(0): /home/bess/bess/core/bessd(_ZN5Queue7RunTaskEP7ContextPN4bess11PacketBatchEPv+0xc7) [0x557a0a5494a7]
    Queue::RunTask(Context*, bess::PacketBatch*, void*) at /home/bess/bess/core/modules/../kmod/../pktbatch.h:54
         51:   // WARNING: this function has no bounds checks and so it's possible to
         52:   // overrun the buffer by calling this. We are not adding bounds check because
         53:   // we want maximum GOFAST.
      -> 54:   void add(Packet *pkt) { pkts_[cnt_++] = pkt; }
         55:   void add(PacketBatch *batch) {
         56:     bess::utils::CopyInlined(pkts_ + cnt_, batch->pkts(),
         57:                              batch->cnt() * sizeof(Packet *));
     (inlined by) Queue::RunTask(Context*, bess::PacketBatch*, void*) at /home/bess/bess/core/modules/queue.cc:222
         219:   bess::Packet *pkt;
         220:   ret = llring_sc_dequeue(queue_, (void **)&pkt);       //queue type: struct llring
         221:   if(!ret){
      -> 222:     batch->add(pkt);
         223:     cnt++;
         224:   }
         225: 
(1): /home/bess/bess/core/bessd(_ZNK4TaskclEP7Context+0x67) [0x557a0a416457]
    Task::operator()(Context*) const at /home/bess/bess/core/task.cc:53
      -> 53:   struct task_result result = module_->RunTask(ctx, &init_batch, arg_);
(2): /home/bess/bess/core/bessd(_ZN4bess16DefaultScheduler12ScheduleLoopEv+0x1d0) [0x557a0a3d33d0]
    bess::DefaultScheduler::ScheduleLoop() at /home/bess/bess/core/scheduler.h:272
      -> 272:       auto ret = (*ctx->task)(ctx);
     (inlined by) bess::DefaultScheduler::ScheduleLoop() at /home/bess/bess/core/scheduler.h:250
      -> 250:       ScheduleOnce(&ctx);
(3): /home/bess/bess/core/bessd(_ZN6Worker3RunEPv+0x201) [0x557a0a3cfc11]
    Worker::Run(void*) at /home/bess/bess/core/worker.cc:323
      -> 323:   scheduler_->ScheduleLoop();
(4): /home/bess/bess/core/bessd(_Z10run_workerPv+0x7c) [0x557a0a3cfeec]
    run_worker(void*) at /home/bess/bess/core/worker.cc:337
      -> 337:   return current_worker.Run(_arg);
(5): /home/bess/bess/core/bessd(+0x1191a6e) [0x557a0b0aea6e]
    execute_native_thread_routine at thread.o:?
(6): /lib/x86_64-linux-gnu/libpthread.so.0(+0x76da) [0x7f162bd9e6da]
    start_thread at /build/glibc-uZu3wS/glibc-2.27/nptl/pthread_create.c:463
        (file/line not available)
(7): /lib/x86_64-linux-gnu/libc.so.6(clone+0x3e) [0x7f162b30d61e]
    clone at /build/glibc-uZu3wS/glibc-2.27/misc/../sysdeps/unix/sysv/linux/x86_64/clone.S:95
        (file/line not available)

My test demo is:

import scapy.all as scapy

eth = scapy.Ether(dst='00:02:15:37:a2:44', src='00:ae:f3:52:aa:d1')
ip = scapy.IP()
udp = scapy.UDP()
payload = 'Hello World'

test_packet = bytes(eth/ip/udp/payload)

bess.add_worker(0, 0)
bess.add_worker(1, 1)

src::Source() \
        -> Rewrite(templates=[test_packet]) \
        -> VLANPush(tci=2) \
        -> queue::Queue() \
        -> Sink()

#queue.set_burst(burst=4)

bess.add_tc('fast', policy='rate_limit',wid=0, resource='bit', limit={'bit': 1000000})
src.attach_task('fast')

bess.add_tc('slow', policy='rate_limit', wid=1, resource='bit', limit={'bit': 1000000})
queue.attach_task('slow')

# To get queue occupancy, issue this command in bessctl once this script is loaded:
# command module queue get_status QueueCommandGetStatusArg

How can I solve this problem, looking forward to your reply :)